Automatic image capture

BACKGROUND OF THE INVENTION

p-0002The present invention relates to image processing, and more particularly to an improved automatic image capture method using an intelligent mobile device, such as a wireless telephone having a camera.

p-0003Conventional image capturing devices, such as cameras and scanners, typically require human or other external intelligence to select the appropriate image for a specific application. While this is sufficient for some image capture systems, many systems require very precise images for processing.

p-0004One such example is the new methodology being used by banking institutions for deposit of checks remotely via an intelligent mobile device, such as a wireless telephone, as exemplified by U.S. Pat. Nos. 7,778,457, 7,949,176 and 7,953,268 entitled “Systems [Methods] for Mobile Image Capture and Processing of Checks [Documents]” by Grigori Nepomniachtchi and assigned to Mitek Systems, Inc. of San Diego, Calif., and further shown in a video at http://www.youtube.com/watch?v=sGD49ybxS2Q. Another video of a method used by USAA Bank is shown at http://www.youtube.com/watch?v=waBQqsSg2NM which describes the USAA Deposit @ Mobile system using an iPhone® mobile device, apparently also using the system/method described in the above patents. With this system/method an appropriate application is downloaded to the iPhone device once the user is qualified by USAA Bank. The user accesses the application on the iPhone device and enters the amount of the check. Then the user uses a camera which is part of the iPhone device to capture an image of both the front and back of the check by, for each instance, lining up the check to be captured within a rectangular graticule or frame on the iPhone viewing screen. Once the images are captured, the user is asked to verify the images and then submit the images to USAA Bank. The captured images of the check are then transmitted via wireless or WiFi to USAA Bank. However it can be tedious and difficult for the user to properly align the image of the check within the rectangular graticule, and motion by the user in taking the picture often results in unusable images. Also in some instances the banking institution may require that the check still be sent to the bank as well since the captured image may not satisfy Federal Reserve Board standards regarding legal replacement checks.

p-0005Co-pending U.S. patent application Ser. No. 12/930,630, filed Jan. 12, 2011 by Mike Coleman entitled “Automatic Image Capture” and assigned to iParse, LLC of Beaverton, Oreg., describes an automatic image capture system that does not require human intervention. A user, having the appropriate application downloaded on the intelligent mobile device, aims the device camera at an object or document to be captured. The device camera starts capturing images of the object, and each image is qualified to determine whether it is in focus and within the camera field of view. Qualified images are stored for further processing, such as alignment by an appropriate perspective transformation to assure they each fill a common reference frame, before transmission to a remote location.

p-0006One disadvantage of the above-described automatic image capture system is that it still takes the user a little time to aim the camera at the object, preferably directly from above. As a result, the user and camera could cast a shadow on the object, making it more difficult to capture a clear image. Preferably the camera should be able to capture a single image, rather than taking a plurality of images and qualifying each one.

p-0007What is desired is an improved automatic image capture system that is easy to use and guides the user to place the camera of the intelligent mobile device in such a position that only a single image needs to be captured.

BRIEF SUMMARY OF THE INVENTION

p-0008Accordingly the present invention provides an improved automatic image capture system for an intelligent mobile device having a camera that guides a user to position the camera so only a single image needs to be automatically captured. Application software residing on the intelligent mobile device guides the user, using a trapezoidal view finder on a display of the intelligent mobile device, to orient the camera with respect to a target document so there is an appropriate pitch and roll angle between the camera and the target document to avoid shadows caused by the camera or user. The image of the target document within the view finder is tested for sufficient brightness and contrast with a background on which the document resides. The user is then guided to move closer or farther from the document so that the document is maximized within the view finder. Then the corners of the document are located, and the camera is checked to see is the displayed document image is focused. The camera may also be checked for stability, since excessive movement may cause blurring. When all of the above conditions are satisfied, the application software automatically has the camera capture the image of the document for post-processing. If any of the above conditions fall out of limits prior to the automatic capture step, the process returns to the step immediately preceding the failed step and repeats the steps until all the conditions are met.

p-0009The objects, advantages and other novel features of the present invention are apparent from the following detailed description when read in conjunction with the appended claims and attached drawing figures.

BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWING

p-0010<figref idrefs="DRAWINGS">FIG. 1</figref> is a plan view of a split view finder display for automatic image capture according to the present invention.

p-0011<figref idrefs="DRAWINGS">FIG. 2</figref> is a plan view of a target view finder display for automatic image capture according to the present invention.

p-0012<figref idrefs="DRAWINGS">FIG. 3</figref> is a plan view of a contrast display for automatic image capture according to the present invention.

p-0013<figref idrefs="DRAWINGS">FIG. 4</figref> is a plan view of an occupancy display for automatic image capture according to the present invention.

p-0014<figref idrefs="DRAWINGS">FIG. 5</figref> is a plan view of a corner detection display with status indicators for automatic image capture according to the present invention.

p-0015<figref idrefs="DRAWINGS">FIG. 6</figref> is a flow diagram view of the application software for automatic image capture according to the present invention.

DETAILED DESCRIPTION OF THE INVENTION

p-0016The system requirements for the improved automatic image capture system described below are:

h-00051. An intelligent mobile device, such as a smart phone, that has a high resolution video camera;

h-00062. An operating system (OS) on the intelligent mobile device that provides access to individual video frame data;

h-00073. A high quality display and graphical library capable of rendering images, graphics and text; and

h-00084. Application software running on the OS that provides the algorithms and guidance to capture an image.

h-0009The intelligent mobile device may also contain motion sensors, illumination and network connectivity.

p-0017A user is guided by the application software to improve the conditions needed to capture a high quality image suitable for submission to a remote location, such as a financial or other business/legal institution as an example. The image must be of high enough quality, properly framed and properly filtered so that the rate of rejected images at the remote location is as low as possible. By combining the user with the application software, the best possible image capture is possible. The proximity of the application software within the intelligent mobile device provides realtime feedback that reduces user frustration, eliminates delay in round-trip submissions to a back-end system at the remote location, and gives immediate indication of success to the user. The result is a much lower error rate than conventional solutions that ship a generic, un-corrected image to the back-end system that performs the processing without user cooperation. This benefits both the user and the remote location institution with higher customer satisfaction and lower cost of operation.

p-0018The user may capture a printed rectangular document, i.e., acquire a centered and cropped image of the document in the form of an electronic image, such as PNG, TIFF, JPEG, etc. files, suitable for transfer across standard internet protocols in the following manner. First, the user places the document on a contrasting, pattern-free surface—generally a dark background since most documents are produced on light material. The user ensures sufficient lighting, and then holds the camera of the intelligent mobile device at a preferred angle with respect to the document to eliminate any shadows. Then the user looks at the device display to see a live video image of the document provided by the camera. The user frames the document within a target view finder displayed on the device, and follows the feedback from the application software to improve camera angle, lighting, contrast, occupancy, orientation, focus and motion stability. When all the conditions are right, the camera automatically captures the image, i.e., takes a picture of the document, and the application software automatically performs post-processing of the image to de-skew and filter the image into a format suitable for reception by the remote location for back-end processing.

p-0019The application software provides a view finder <b>10</b>, as shown in <figref idrefs="DRAWINGS">FIG. 1</figref>, to guide the user to hold the camera at a preferred angle, or pitch, with respect to vertical. Conventionally the camera is held at a zero angle with respect to vertical. However the preferred angle is offset from vertical, such as between fifteen (15) and twenty (20) degrees relative to vertical with respect to the document to be captured. By holding the camera tipped back, the illumination source is much less likely to cast a shadow of the camera and user's hands onto the document. The view finder <b>10</b> has a trapezoidal shape, as opposed to the conventional rectangular shape, to match the perspective of the document as seen when looking through the camera when it is tipped back by the preferred angle so the user can naturally align the document's sides and top with the perspective view finder. The view finder <b>10</b> is static within the display. A secondary view finder <b>12</b> or corresponding shape to the static view finder <b>10</b> is provided and moves with the movement of the camera relative to the document. This forms a split view finder together with the static view finder <b>10</b>. The user is guided to line up the two trapezoidal view finders <b>10</b>, <b>12</b>, i.e., move the camera until the secondary view finder is superimposed on the static view finder, which results in the correct pitch and roll (zero degrees). The outlines of the two view finders <b>10</b>, <b>12</b> may be an appropriate negative color, such as cyan or red, so long as the camera is misaligned with the document. When the two view finders <b>10</b>, <b>12</b> are sufficiently superimposed on each other, indicative of proper alignment, the view finders may change to a positive color, such as green, to indicate successful alignment in pitch and roll.

p-0020Alternatively as shown in <figref idrefs="DRAWINGS">FIG. 2</figref>, a cross-hair <b>14</b> may be provided that moves with the pitch and roll of the camera while a small target box <b>16</b> is centered statically within the view finder <b>10</b>. The user moves the camera to move the cross-hair <b>14</b> into the target box <b>16</b>, assuring proper camera alignment with the document. When the camera is misaligned with respect to the document, the target box <b>16</b> and cross-hair <b>14</b> may have a distinctive color, such as cyan, and the live video feed is darkened significantly. Also the trapezoidal view finder <b>10</b> may turn from a solid outline to a dashed outline with some guidance text shown on the display, such as “Tilt camera to align cross-hair within box.” When the camera and document are aligned, the video feed of the document is restored to its normal brightness, and the target box <b>16</b> and cross-hair <b>14</b> change to a positive color, such as green. The target box <b>16</b> and cross-hair <b>14</b> remain the positive color as the rest of the conditions required for image capture are achieved.

p-0021A warning icon is displayed when there is low light on the document to be captured. Brightness may be measured by averaging pixel values that approximate luminance in several areas within the trapezoidal view finder, i.e., taking the average of the pixel values that reside within each sub-area of the document image. This avoids acceptance of strong shadows that might otherwise increase error rates in the post-processing steps. Again, text might be displayed, if the brightness is insufficient, to instruct the user to provide more light for the document.

p-0022As shown in <figref idrefs="DRAWINGS">FIG. 3</figref>, a patterned or colored area <b>18</b> outside the view finder <b>10</b> is used to indicate poor contrast. The application software compares pixels inside and outside the document image <b>20</b> in different areas to determine if sufficient brightness contrast exists. A warning icon and text also suggests corrective action to increase contrast, such as changing the background against which the document is placed. Selected pixels <b>18</b>, <b>22</b> both within and outside document image <b>20</b> are summed separately, a threshold is created using a center value between the two sums, and the average within the document to that outside the document is compared. A low difference indicates poor contrast.

p-0023With camera orientation and brightness/contrast being satisfactory, the user is then guided to fill the view finder <b>10</b> with the document <b>20</b>, as shown in <figref idrefs="DRAWINGS">FIG. 4</figref>. Using as much of the view finder <b>10</b> as possible reduces noise in the final image. Arrows <b>24</b> or other indicators on the display guide the user to move the camera closer or farther back to align the document image <b>20</b> optimally within the trapezoid view finder <b>10</b>. The arrows <b>24</b> may be colored, such as red. To determine the state of each arrow <b>24</b>, a first order differential is applied at each of four specific locations and determines if the difference is high enough. This is similar to contrast, but is measured separately at the four points shown by the arrows <b>24</b>. Each point of measure has an equivalent warning arrow <b>24</b> that illuminates when the differential is too low. The four areas are: (1) each side at the center row of the view finder <b>10</b>—this is a horizontal differential that is maximized when the edge of the document image <b>20</b> is within a given offset with the view finder; and (2) top left (set in) and top right (set in)—these are vertical differentials that are maximized when the top edge of the document image is aligned with the top edge of the view finder, minus a few rows. Using two arrows <b>24</b> at the top minimizes any rotation of the document image <b>20</b>, which enables corner detection in the next step. There is no need for bottom arrows, which allows for documents having different aspect ratios, such as business/personal checks and letter/legal sized pages, to be imaged and captured.

p-0024As an alternative to using arrows to indicate occupancy, the user may be guided to fill the view finder area by animated movements of the view finder that suggest actions the user should take to improve occupancy. For example, when the camera is too close to the document and the document image exceeds the bounds of the view finder, the view finder may show a repeating expanding motion—the view finder animates a series of increased sizes to suggest pulling back of the camera by the user. Also for example, when the camera is too far away from the document such that the document image is occupying too small an area in the view finder, the view finder may animate a shrinking motion. This would suggest to the user that the camera should be moved closer. Similar animations may be used to suggest moving the camera left or right, up or down, etc. The distances from the edge of the document image to the edges of the view finder would provide the information as to the required direction of movement, i.e., the distance from the top of the document image to the top of the view finder may be significantly greater than the distance from the bottom of the document image to the bottom of the view finder so the camera should be moved up to center the document image within the view finder, as an example.

p-0025For edge and corner detection as shown in <figref idrefs="DRAWINGS">FIG. 5</figref>, a first order differential is applied in eight critical areas to determine four lines <b>26</b>L, R, T, B that make up the outline of the document image <b>20</b>. The intersections of the four lines <b>26</b>L, R, T, B determine the corners <b>28</b>. This assures that a rectangle is present within the optimal area of the view finder <b>10</b>. This step is very fast. An audible “beep” may be provided to help give the user the feeling of progress each time the four corners <b>28</b> are located.

p-0026More specifically in each corner the application software tries to find two points as the upper/lower points <b>30</b> and the outer points <b>32</b>. This leads to the four lines: upper, lower, left and right. The left and right lines <b>26</b>L, R are found by connecting the two outer points <b>32</b>, and the upper and lower lines <b>26</b>T, B are found by connecting the two upper and lower points <b>30</b> respectively. To find each point <b>30</b>, <b>32</b>, a first order differential is used starting, for example, ten percent (10%) outside the view finder <b>10</b> and moving inwards until ten or twenty percent (10-20%) inside the view finder. The point at which the maximum luminance change occurs, i.e., the transition between the background and the document image <b>20</b>, is considered the edge of the document. This process is repeated for each of the eight points <b>30</b>, <b>32</b>. Finally projections of the lines <b>26</b>L, R, T, B are intersected to produce the corners <b>28</b>. The four corners are used in post-processing to improve performance and accuracy of the final document bounds detection.

p-0027The application software provides a warning while the camera is automatically focusing to give the user feedback that the capture process is still proceeding.

p-0028The final condition, if necessary, is stability, where motion sensors are used to measure the physical motion of the intelligent mobile device. This condition may not be necessary for image capture, but may be included in the event there is a possibility of excessive motion during the capturing of the final image. Excessive motion may cause blurring and increase the error rate in post-processing. Accelerometer data is acquired from the intelligent mobile device and filtered using a lowpass filter. High vibration causes blurs, but low vibration does not. Excessive low frequency movement, such as moving closer to the document, may trigger an “out-of-focus” condition that would return the application software to the focus step. If high frequency motion, such as camera vibration, is detected, a warning icon and corrective help text may be provided on the device display until stability is achieved. Once stable, all of the conditions required to capture the image have been set, and the application software automatically captures the image of the document and proceeds to the post-processing and filtering, as described in the afore-mentioned pending U.S. patent application Ser. No. 12/930,630.

p-0029All of these steps are combined into a framework that guides the user through the required conditions in a natural order, as shown in <figref idrefs="DRAWINGS">FIG. 6</figref>. When the application software starts, the first step <b>40</b> is orientation of the camera with respect to the target document. The next step <b>42</b> is to check for sufficient brightness and contrast for the target document. At this point occupancy is checked as the next step <b>44</b> to assure the document image fills the view finder. The final step <b>46</b> is to detect the corners of the document image. An additional <b>48</b>, as necessary or desirable, is to check for excessive motion to assure sufficient stability of the camera with respect to the target document to prevent blurring. Once all the conditions of the previous steps are satisfied, then the automatic image capture <b>50</b> of the target document occurs and the captured image is passed to post-processing algorithms before being transmitted to the remote location. In the event that any of the conditions of prior steps fail before the image capture step <b>50</b>, the process returns to the immediately preceding step of the failed condition and repeats the ensuing steps.

p-0030These steps may include other conditions not listed above which have already been described in the pending U.S. patent application Ser. No. 12/930,630. The user is only provided with the lowest “out of bounds” conditions that require corrective action. There is no sense trying to fix “focus” if there is not enough “brightness.” Likewise view angle adjustment is the first step <b>40</b> because moving the camera around changes all of the other conditions. A progress meter <b>34</b>, shown in <figref idrefs="DRAWINGS">FIG. 5</figref> for example, that has a simple color indicator for each step may be presented at the top of the view finder <b>10</b> to give the user feedback as to where they are in the process of getting a good image capture of the document.

p-0031By combining the user's eyes and problem solving ability with supportive image processing algorithms, the application software guides the user to capture a high quality image of the document sufficient to transmit to the remote location and be accepted at a very low error rate, giving the user an excellent experience while providing low cost of business.

p-0032Thus the present invention provides an improved automatic image capture of a target document using an intelligent mobile device having a camera by providing application software that operates interactively with a user to assure that a series of conditions are achieved, starting with camera tilt relative to the target document, before the application software causes the camera to automatically capture the image of the target document for post-processing and subsequent transmission to a remote location.